Thank you! The conclusion is quite good, like use spark as ETL and Presto (Trino) for analytical queries but the article looks very outdated.
Spark is not about RDDs. Today the most usage of Spark is via DataFrame API. And it is not just syntax. The Catalyst itslef provide a lot of performance optimizations, like predicate pushdown on the level of orc/parquet reading, automatic skew joins detection, prunning, etc.
Also Presto in this case should be called as Trino because there was a rebranding in 2020
As open source replacement of google drive I would prefer to use Nextcloud